Data versus decision fusion for classification in sensor networks
A. D'Costa, A.M. Sayeed · 2003
Sensor networks provide virtual snapshots of the physical world via distributed wireless nodes that can sense in different modalities, such as acoustic and seismic. Classification of objects moving through the sensorfield is an important application that requires collaborative signal processing (CSP) between nodes. Two main forms of CSP are possible. Data fusion - exchange of low dimensional feature vectors - is needed between correlated nodes, in general, for optimal performance. Decision fusion - er- change of likelihood values - is sufficient between indepen- dent nodes. Decision fusion is generally preferable due to its lower communication burden. We study CSP of multiple node measurements in the context of single target classif- cation. Each measurement is modeled as a Gaussian sig- nal vector (corresponding to the taaet class) corrupted by additive white Gaussian noise. The measurements are par- titioned into groups. The signal components within each group are perfectly correlated whereas they vary indepen- dently between groups. Three classifiers are compared: the optimal man'mum likelihood classifier; a data averaging classifier that treats all measurements as correlated, and a decisionfusion classifier that treats them all as indepen- dent. Analytical and numerical results based on real data are provided to compare the performance of the three CSP classifiers. Our results indicate that the sub-optimal deci- sionfusion classifier; that is most attractive in the context of sensor networks, is also a robust choice from a decision theoretic viewpoint.